Flight dynamic management method and system

Abstract

The invention provides a flight dynamic management method and system, and the method comprises the steps: selecting a constructed flight result, and obtaining all flight routes in the flight result; extracting related attribute information of all flight routes in the flight result; based on the flight related attribute information, the probability that all the routes are selected in the flight result is calculated, if the probability is higher than an expected value, the corresponding flight route is stored, and otherwise, the corresponding flight route is put into a recycle bin; based on the number of the stored flight results, sorting the flight results from large to small according to the probability of being selected, and selecting the flight route which is sorted in the front as the optimal flight route; the optimal flight route is stored, the optimal flight route data is fed back, and if the optimal flight route data does not meet the user requirement, the result with the high probability is selected from the recycle bin for supplementation, so that the richness of the flight route is met. According to the method, the optimal flight result can be selected from mass flight data, and the requirements of different user groups are met.

Description

Technical Field

[0001] The invention belongs to the technical field of flight query recommendation, and in particular relates to a flight dynamic management method and system. Background Art

[0002] Since the fare search system faces the global market, users often request a large number of solutions, including direct flights, single transfers and multiple transfers. In the actual environment, it is unrealistic to feed back all flight results to users, both from the perspective of system response time and user experience. At the same time, users of the fare search system include agents from various channels, airlines, etc. Different user groups have different demands and perceptions of flights, and there will be various different situations.

[0003] In the prior art, the application number is 201910321627.X, and the name of the invention patent is: A method for intelligently recommending air tickets. A method for intelligently recommending air tickets is disclosed. The method scores flights based on multiple factors, recommends flight data to users according to the scores, and recommends only a small amount of flight information so that users can choose the air tickets they want, which can effectively eliminate the user's choice phobia. In this patent, flights are scored mainly based on the flight fare and comfort level, and the weight of each flight is determined based on the user's historical search data, such as whether the flight has been searched, whether it has been added to the shopping cart, whether it is a popular route, and whether the departure point of the flight is a common departure point or whether it is consistent with the user's current location. Multiple factors such as recommending flight information to solve the user's choice phobia. However, this method only stands from the user's perspective and only considers the user's needs, and does not take into account the unique attributes of flights in the aviation industry, such as the number of transfer points, flight time, airline type, aircraft model, etc., which will cause the recommended flight information to bring uncertainty to the user, that is, the recommended flight information is not the optimal flight information.

[0004] Therefore, how to select the best flight results from massive flight data to meet the needs of different user groups as much as possible is the main challenge currently facing the fare search system. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a flight dynamic management method and system, which introduces the MNL model (passenger selection model) in the field of machine learning and artificial intelligence into the fare search system, screens and filters the results of direct flights and connecting flights, so that the most ideal and effective optimal flight results can be displayed to users in the fare search system.

[0006] The present invention is achieved through the following technical solutions:

[0007] In one aspect, the present invention provides a flight dynamic management method, which specifically comprises the following steps:

[0008] Step S1: Select a completed flight result and obtain all flight routes in the flight result;

[0009] Step S2: extracting relevant attribute information of all flight routes in the flight result;

[0010] Step S3: Based on the flight-related attribute information, the probability of all routes in the flight result being selected is calculated. If the probability is higher than the expected value, the corresponding flight result is stored; otherwise, the corresponding flight result is placed in a recycle bin;

[0011] Step S4: based on the number of stored flight results, sort them from large to small according to the probability of being selected, and select the flight route with the highest ranking as the optimal flight route;

[0012] Step S5: Store the optimal flight route and feedback the optimal flight route data. When the feedback optimal flight route data does not meet the user's needs, select flight route data with a higher probability from the recycle bin to supplement it to meet the richness of the flight route.

[0013] Furthermore, in step S1, the selection of a completed flight result is specifically all scheduled flight routes from the starting point to the destination, including direct flights and transfer routes.

[0014] Furthermore, in step S2, the flight attribute information includes: number of transfer points, airline type, flight time, price, AV (airline company), departure date, departure time, arrival time and aircraft type. The flight attribute information is determined by the passenger selection model optimization module or by the user's self-selection.

[0015] In one aspect, the present invention provides a flight dynamic management system, which includes: a flight creation module, a flight management module, a data collection module, a passenger selection model optimization module, a manual intervention module, a record update module and a data storage module, wherein:

[0016] The flight creation module is used to generate flight results;

[0017] The data collection module is used to collect data sources from external systems and organize the data sources into a data set in CSV format as a data training set for the passenger selection model optimization module;

[0018] The passenger selection model optimization module is used to build the model, and the data set collected and sorted by the data collection module is periodically trained and fitted in the built model to obtain model parameters, and the model parameters are tuned to obtain the final modeling data;

[0019] The flight management module dynamically screens and filters the flight results generated by the flight creation module based on the modeling data of the passenger selection model optimization module;

[0020] The manual intervention module is used to input screening conditions or formulate rules for manual intervention when abnormal data or results that do not meet user requirements appear in the passenger selection model optimization module;

[0021] The record update module is used to record and update the flight data information dynamically screened and filtered by the flight management module, and store the flight data information in the data storage module.

[0022] Among them, in the flight creation module, the flight creation module generates flight results including, based on the departure and destination of the flight, obtaining all flight routes from the departure to the destination.

[0023] Among them, in the data collection module, the data collection module is used to collect data sources from the external system and organize the data sources into a data set in CSV format as a data training set for the passenger selection model optimization module, and also includes:

[0024] The data sources collected by the data collection module from the external system on a regular basis are classified, sorted and aligned according to the data collection points (DCPs), and the attribute information of the data sources is extracted;

[0025] Based on the extracted attribute information of the data source, the data source is organized into a dataset in CSV format.

[0026] Furthermore, the external system includes a fare calculation system and an airline front-end system; the data source includes PNR (passenger reservation record) data, historical price verification data, historical fare search data, and historical ticket data; the attribute information of the data source includes the number of transfer points, airline type, flight time, price, AV (airline company), departure date, departure time, arrival time, and aircraft type;

[0027] Furthermore, when the data source is organized into a CSV format data set based on the attribute information of the extracted data source, it is also necessary to quantify the airline type, departure date, departure time, arrival time and aircraft type attribute information so that the passenger selection model optimization module can directly use it;

[0028] Furthermore, the quantification of the airline type, departure date, departure time, arrival time and aircraft model data includes:

[0029] Quantify the airline type into 0, 1, 2 according to the alliance it belongs to;

[0030] The departure date is quantified into 1-7 (day of the week) according to the 7 days of a week;

[0031] Quantify the departure and arrival times as the number of minutes from 0 o'clock (e.g. 1 o'clock can be quantified as 60 minutes);

[0032] The model is quantified into 0 or 1 according to its width or narrowness.

[0033] Among them, in the passenger selection model optimization module, the passenger selection model optimization module is used to build the model, and the data set collected and sorted by the data collection module is periodically trained and fitted in the built model to obtain model parameters, and the model parameters are tuned to obtain the final modeling data; specifically including:

[0034] Use LogisticRegression in pytorch or sklearn as the modeling framework for modeling;

[0035] After the modeling is completed, the data set is trained and fitted to obtain model parameters;

[0036] Based on the obtained model parameters, new data are input for prediction and evaluation to obtain reliability evaluation indicators. The model parameters are tuned according to the reliability evaluation indicators to obtain the final modeling data.

[0037] Furthermore, when the LogisticRegression in pytorch or sklearn is used as a modeling framework for modeling, the modeling is performed in a weekly or monthly manner, while ensuring the decoupling of the framework and maintaining the flexibility of the framework.

[0038] Compared with the prior art, the present invention has the following advantages:

[0039] The present invention applies the MNL passenger selection model optimization module to the fare search system, performs a comprehensive evaluation on multiple dimensions of flight results, calculates the selection probability of flight results, and combines the unique attributes of flights while considering different user needs, so that the fare search system can select the best result from a large number of flight results and complete the filtering and screening of flight results; while ensuring the quality of the results, the system response speed is improved, thereby improving user satisfaction.

[0040] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 : A flow chart of a flight dynamic management method;

[0043] Figure 2 : A schematic diagram of the structure of a flight dynamic management system;

[0044] Figure 3 :PNR (Passenger Reservation Record) data example diagram;

[0045] Figure 4 : Example chart of ticket sales data;

[0046] Figure 5 : Passenger selection model architecture diagram;

[0047] Figure 6 : Sample diagram of the data set;

[0048] Figure 7 : Sample diagram of the processed data set;

[0049] Figure 8 : Example diagram of flight sorting process. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] In one embodiment, see Figure 1 , provides a flight dynamic management method, the method specifically comprising:

[0052] Step S1: Select a completed flight result and obtain all flight routes in the flight result;

[0053] Furthermore, the flight results are all scheduled flight routes from the origin to the destination, including direct flights and transfer routes.

[0054] Step S2: extracting relevant attribute information of all flight routes in the flight result;

[0055] Furthermore, the relevant attribute information of the flight route includes: number of transfer points, type of airline, flight time, price, AV (airline), departure date, departure time, arrival time and aircraft type.

[0056] Step S3: Based on the flight-related attributes, the probability of all routes in the flight result being selected is calculated. If the probability is higher than the expected value, the corresponding flight route is stored; otherwise, the corresponding flight route is put into the recycle bin;

[0057] Step S4: based on the number of stored flight results, sort them from large to small according to the probability of being selected, and select the flight route with the highest ranking as the optimal flight route;

[0058] Step S5: Store the optimal flight route and feedback the optimal flight route data. When the feedback optimal flight route data does not meet the user's needs, select flight route data with a higher probability from the recycle bin to supplement it to meet the richness of the flight route.

[0059] In one embodiment, see Figure 2 to Figure 7 , provides a flight dynamic management system, such as Figure 2 As shown, the system includes: a flight creation module, a flight management module, a data collection module, a passenger selection model optimization module, a manual intervention module, a record update module and a data storage module, wherein:

[0060] The flight creation module is used to generate flight results;

[0061] Furthermore, the flight creation module generates flight results including, based on the departure place and the destination, obtaining all flight routes from the departure place to the destination.

[0062] The data collection module is used to collect data sources from external systems and organize the data sources into a data set in CSV format as a data training set for the passenger selection model optimization module;

[0063] Furthermore, the data collection module also includes:

[0064] The data sources collected by the data collection module from the external system on a regular basis are classified, sorted and aligned according to the data collection points (DCPs), and the attribute information of the data sources is extracted;

[0065] Based on the extracted attribute information of the data source, the data source is organized into a dataset in CSV format.

[0066] Furthermore, the external system includes a fare calculation system and an airline front-end system; the data source includes PNR (passenger reservation record) data, historical price verification data, historical fare search data, and historical ticket data; the attribute information of the data source includes the number of transfer points, airline type, flight time, price, AV (airline company), departure date, departure time, arrival time, and aircraft type;

[0067] Furthermore, when the data source is organized into a CSV format data set based on the attribute information of the extracted data source, it is also necessary to quantify the airline type, departure date, departure time, arrival time and aircraft type attribute information so that the passenger selection model optimization module can directly use it;

[0068] Furthermore, the quantification of the airline type, departure date, departure time, arrival time and aircraft model data includes:

[0069] Quantify the airline type into 0, 1, 2 according to the alliance it belongs to;

[0070] The departure date is quantified into 1-7 (day of the week) according to the 7 days of a week;

[0071] Quantify the departure and arrival times as the number of minutes from 0 o'clock (e.g. 1 o'clock can be quantified as 60 minutes);

[0072] The model is quantified into 0 or 1 according to its width or narrowness.

[0073] The passenger selection model optimization module is used to build the model, and the data set collected and sorted by the data collection module is periodically trained and fitted in the built model to obtain model parameters, and the model parameters are tuned to obtain the final modeling data;

[0074] Furthermore, in the passenger selection model optimization module, it also includes:

[0075] Use LogisticRegression in pytorch or sklearn as the modeling framework for modeling;

[0076] After the modeling is completed, the data set is trained and fitted to obtain model parameters;

[0077] Based on the obtained model parameters, new data are input for prediction and evaluation to obtain reliability evaluation indicators. The model parameters are tuned according to the reliability evaluation indicators to obtain the final modeling data.

[0078] Furthermore, when the LogisticRegression in pytorch or sklearn is used as a modeling framework for modeling, the modeling is performed in a weekly or monthly manner, while ensuring the decoupling of the framework and maintaining the flexibility of the framework.

[0079] The flight management module dynamically screens and filters the flight results generated by the flight creation module based on the modeling data of the passenger selection model optimization module;

[0080] The manual intervention module is used to input screening conditions or formulate rules for manual intervention when abnormal data or results that do not meet user requirements appear in the passenger selection model optimization module;

[0081] The record update module is used to record and update the flight data information dynamically screened and filtered by the flight management module, and store the flight data information in the data storage module.

[0082] Further, in this embodiment, please refer to Figure 5 , for further explanation of the tourist selection model optimization module; please refer to Figure 3 and Figure 4 Taking the PNR (passenger reservation record) data and historical ticket data in the data source collected by the data collection module as an example, the PNR (passenger reservation record) data and historical ticket data are obtained (see Figure 6 ) and extract the attribute information from the PNR (passenger reservation record) data and historical ticket data, and organize them into a CSV format data set (see Figure 7 ); the passenger selection model optimization module is modeled by using LogisticRegression in pytorch or sklearn as a modeling framework, and then training and fitting are performed in the constructed model based on the data set to obtain model parameters. Based on the obtained model parameters, new data is input for prediction and evaluation to obtain reliability evaluation indicators. The model parameters are tuned according to the reliability evaluation indicators to obtain the final modeling data.

[0083] In one embodiment, this embodiment mainly predicts the selection probability of each flight in the fare search system based on the modeling data of the passenger selection model optimization module. Specifically:

[0084] Assume that a user searches for flight results from Beijing to Frankfurt on May 30, 2024. After flight calculation and preliminary fare screening, the fare search system calculates a total of f1-f5 flight results, as shown in Table 1.

[0085] Table 1: Results of flights screened by the fare search system.

[0086]

[0087] After obtaining the 5 flight results in Table 1, the results will be passed to the passenger selection model for selection probability prediction. Figure 8 After the selection probability prediction is completed, the final sorting results are shown in Table 2: The final sorting results of flights are shown in the table below.

[0088] Table 2: Final flight sorting results display table

[0089] Flight Results Number of transfer points Departure time Flight time price model Selection probability f4 0 9:00 10 9000 919 60％ f5 0 11:00 10.5 8900 919 50％ f1 1 8:00 13 7000 310 / 320 15％ f2 1 18:00 15 6500 310 10％ f3 2 7:00 24 4000 320 5％

[0090] As shown in Table 2, after calculating the flight results f1 to f5, the fare search system will extract the attributes of multiple dimensions such as the number of transfer points and take-off time of the flight results in turn, and calculate the selection probability based on these attributes; if the user only needs 3 results, f4, f5 and f1 will be displayed to the user, and f2 and f3 will be filtered out.

[0091] Furthermore, if the sorting results in Table 2 do not meet the user's expectations, the user can manually enter the screening conditions or formulate screening rules through the manual intervention module for manual intervention. At this time, the passenger selection model optimization module will re-model and calculate based on the data set based on the screening conditions manually entered or the formulated screening rules in the manual intervention module to obtain the final sorting results.

[0092] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A flight dynamic management method, characterized in that: The method comprises: Select a completed flight result and obtain all flight routes in the flight result; Extract relevant attribute information of all flight routes in the flight results; Based on the flight-related attribute information, the probability of all routes in the flight result being selected is calculated. If the probability is higher than the expected value, the corresponding flight result is stored, otherwise the corresponding flight result is put into the recycle bin; Based on the number of stored flight results, the flight routes are sorted from large to small according to the probability of being selected, and the flight routes with the highest ranking are selected as the optimal flight routes; The optimal flight routes are stored and the optimal flight route data is fed back. When the fed back optimal flight route data does not meet the user's needs, the results with higher probability are selected from the recycle bin to supplement them to meet the richness of the flight routes.

2. A flight dynamic management method according to claim 1, characterized in that: The selected constructed flight result specifically refers to all scheduled flight routes from the starting point to the destination, including direct flights and transfer routes.

3. A flight dynamic management method according to claim 1, characterized in that: The relevant attribute information of the flight route includes the number of transfer points, airline type, flight time, price, AV, departure date, departure time, arrival time and aircraft type.

4. A flight dynamic management system, characterized by: It includes flight creation module, flight management module, data collection module, passenger selection model optimization module, manual intervention module, record update module and data storage module, among which: The flight creation module is used to generate flight results; The data collection module is used to collect data sources from external systems and organize the data sources into a data set in CSV format as a data training set for the passenger selection model optimization module; The passenger selection model optimization module is used to build the model, and the data set collected and sorted by the data collection module is periodically trained and fitted in the built model to obtain model parameters, and the model parameters are tuned to obtain the final modeling data; The flight management module dynamically screens and filters the flight results generated by the flight creation module based on the modeling data of the passenger selection model optimization module; The manual intervention module is used to input screening conditions or formulate rules for manual intervention when abnormal data or results that do not meet user requirements appear in the passenger selection model optimization module; The record update module is used to record and update the flight data information dynamically screened and filtered by the flight management module, and store the flight data information in the data storage module.

5. A flight dynamic management system as claimed in claim 4, characterized in that: The data collection module also includes: The data sources regularly collected by the data collection module from the external system are classified, sorted and aligned according to the data collection points, and the attribute information of the data sources is extracted; Based on the extracted attribute information of the data source, the data source is organized into a dataset in CSV format.

6. A flight dynamic management system as claimed in claim 5, characterized in that: The external systems include the freight rate calculation system and the airline front-end system; The data sources include PNR data, historical price verification data, historical fare search data and historical ticket data; The attribute information of the data source includes the number of transfer points, airline type, flight time, price, AV, departure date, departure time, arrival time and aircraft type.

7. A flight dynamic management system according to claim 6, characterized in that: When the data source is organized into a data set in CSV format based on the attribute information of the extracted data source, it is also necessary to quantify the airline type, departure date, departure time, arrival time and aircraft model attribute information to facilitate direct use by the passenger selection model optimization module.

8. A flight dynamic management system as claimed in claim 7, characterized in that: The quantification of the airline type, departure date, departure time, arrival time and aircraft model data includes: Quantify the airline type into 0, 1, 2 according to the alliance it belongs to; The departure date is quantified into 1-7 according to the 7 days of a week, indicating the day of the week; Quantify the departure and arrival times as the number of minutes away from 0 o'clock; The model is quantified into 0 or 1 according to its width or narrowness.

9. A flight dynamic management system as claimed in claim 4, characterized in that: The passenger selection model optimization module also includes: Use LogisticRegression in pytorch or sklearn as the modeling framework for modeling; After the modeling is completed, the data set is trained and fitted to obtain model parameters; Based on the obtained model parameters, new data are input for prediction and evaluation to obtain reliability evaluation indicators. The model parameters are tuned according to the reliability evaluation indicators to obtain the final modeling data.

10. A flight dynamic management system according to claim 9, characterized in that: When using LogisticRegression in pytorch or sklearn as a modeling framework for modeling, the modeling is performed in a weekly or monthly manner, while ensuring the decoupling of the framework and maintaining the flexibility of the framework.

Images